1 Finnish Health Information on the Semantic Web 30 November 2007 Osma Suominen, Kim Viljanen, Eero Hyvönen Semantic Computing Research Group (SeCo) Helsinki University of Technology (TKK), Laboratory of Media Technology and University of Helsinki, Department of Computer Science HealthFinland (Tervesuomi)

2 Outline of Talk Goals of HealthFinland Challenges Content creation KOS in HealthFinland User interface and navigation –demonstration of portal Evaluation of results

3 HealthFinland portal national health promotion portal for citizens [1] SW technologies: ontologies, metadata... –content aggregated from many websites into a single portal diet, exercise, (non)smoking, healthy living... [1] Hyvönen et al: HealthFinland - Finnish Health Information on the Semantic Web. Proc. ISWC 2007, Busan, Korea, Nov 2007

4 Goals of HealthFinland: Citizens Global view to health information from different organizations Aggregated view to information from different organizations Semantic search and browsing –faceted browsing using user-centric categorizations –string search based on ontologies –semantic recommendations of related information

5 Goals of HealthFinland: Publishers Rationalizing content creation by eliminating redundancy Enriching content with other providers’ content Automatic and dynamic content linking Reusing global services cost-efficiently as Web 2.0 mash-ups Using centralized ontology services for indexing

6 Challenges create a compelling user experience for the general public –usability is not just a surface feature! –solve actual problems users have – in an intuitive way tri-lingual portal and KOS –Finnish, Swedish, English interoperable –with existing KOS –legacy metadata, other document repositories and portals –current and future content management systems need to gain a critical mass of information, publishers, users need processes for content creation, quality control, dealing with problems, KOS updates and maintenance

7 Content Creation Tools Content harvester collects metadata from HTML pages and RDF sources Metadata validator with feedback reports SAHA Annotation editor ONKI Ontology Server

8 Metadata Schema in HealthFinland

9 KOS in HealthFinland YSO >20000 concepts fi, en, sv OWL ontology created by FinnONTO (from YSA) General knowledge MeSH >20000 concepts en, fi, sv … SKOS thesaurus created by US NLM Medical domain (for experts) HPMULTI 1200 concepts en, fi, sv … SKOS thesaurus created by EU project Health promotion specific Source KOS + some organization-specific thesauri (TTL, STAMETA, Suomi.fi …) HealthFinland Health Promotion Ontology automatic term-based mapping manual ontology engineering YSO 10% of MeSH HPMULTI (all stored and maintained on Onki ontology server)

10 User interface and navigation Problem: complex ontologies not suitable for human consumption –made for a different purpose (e.g. indexing scientific articles) –expert terminology –unintuitive hierarchies and groupings –everything in a big hierarchy – not facet-based need to build navigation structures that users of the site will understand [2] solved using an approach [3] based on card sorting [2] Pollit: The key role of classification and indexing in view-based searching. Technical report. University of Huddersfield, UK (1998) [3] Suominen et al: User-centric faceted search for semantic portals. Proc. ESWC 2007, Innsbruck, Austria, Jun

11 Example of problem My 70 year old aunt is depressed. How can I find information about mental illnesses in elderly people?

12 Demonstration of HealthFinland Live!

13 Solution approach faceted browsing very useful and usable in earlier portals –Flamenco [4], SWED [5], MuseumFinland [6]... so, need to build intuitive facets and categories for the site to find out what works for users, do real user research [7] when you know what works for users, handle the technicalities –mapping user facets to ontologies [4] Hearst et al: Finding the flow in web site search. CACM 45(9) 2002 [5] [6] Hyvönen et al: MuseumFinland – Finnish museums on the Semantic Web. Journal of Web Semantics 3(2) 2005 [7] Rugg & McGeorge: The sorting techniques: a tutorial paper on card sorts, picture sorts and item sorts. Expert systems 14(2)

14 Card sorting method to find out how users conceptualize the information space (i.e. how they group things in their head) often used to build website navigation (information architecture) –what should the main sections of the site be –how should they be named –what things go where –what things belong together idea: print a stack of cards with names of documents, let users sort them into piles and give names to the piles –make notes repeat with several users, try to find common patterns raw output: sets of labeled piles of cards easy to do, very effective, enlightening for a designer

15

16 Facets and categories

17 Mapping facets to ontology each category contains a set of concepts category taxonomy represented using SKOS vocabulary –labels: skos:prefLabel –hierarchy: skos:broader mappings to ontology using SKOS Mapping –skosmap: exactMatch, skosmap:narrowMatch result: categories contain documents –annotated with concepts contained in the category »or subconcepts (rdfs:subClassOf or skos:broader) –from all subcategories, recursively »creating a proper subsumption hierarchy

18 Example mappings topic:weight_control topic:losing_weight topic:nutrition_food topc:diet Weight control Losing weight Nutrition & Food Diet skos:broader skos:prefLabel mesh:D skosmap:narrowMatch Energy Intake mesh:D Body Weight skos:prefLabel mesh:D Weight Loss skosmap:narrowMatch Caloric Restriction mesh:D skos:broader Ontological conceptsFacet categories skos:prefLabel

19 Evaluation closed card sorting session to test intuitiveness of facets –given the final facets, do test subjects place a set of concepts in the intended categories? –promising results (but only 2 subjects) review of facets by domain experts: some problems corrected –too much lumping gives wrong message to users –omission of important topics prototype portal to test whether the approach can actually be implemented currently performing 3 rd phase user testing with prototype –so far the problems found have not concerned the categorizations –more user tests under way (in the following 2 weeks) –results will be published in academic conferences, journals etc. prototype will go live in early 2008, production use in

20 Thanks! ?! Hyvönen et al: HealthFinland - Finnish Health Information on the Semantic Web. Proc. ISWC 2007, Busan, Korea, Nov 2007 Suominen et al: User-centric faceted search for semantic portals. Proc. ESWC 2007, Innsbruck, Austria, Jun 2007